Trademark image retrieval by integrating shape with texture feature

Trademark image retrieval is important to protect the copyright of registered trademark. A trademark play a significant role to distinguish the symbol or indicator used by an individual, commercial organization, or other authorized entity to recognize that the products or services to clients with which the trademark appears initiate from a unique source. However, measuring perceptual similarity and defining appropriate similarity measure between trademark images remain a challenging task. In this paper, we propose a method for trademark image retrieval which is based on shape and texture features of trademark images. Zernike moment and curvelet transform is used as shape and texture feature respectively. A weighted average distance is used for similarity measure. Results illustrate that combining shape feature with curvelet based texture feature performs well in precision.